NASA Benchmarks the New G5 Powermac

sockit2me9000 writes "Well NASA's Langley Research Center recently benchmarked the new G5 dual 2ghz Powermac against a dual 1ghz Xserve, a dual 1.25 ghz Powermac, a Pentium4 2 ghz, and a Pentium4 2.66 ghz. To make things fair, the second processor in the G5 was switched off, as well as the other dual sysytems. Then, they all ran Jet3d. Even with un-optimized code and one processor, the G5 performance is impressive."

Comments

Not that intresting

[autopr0n]

According to some of the comments on this board, the intel fortran compiler is about 50% faster then the portland compiler that was used. If the intel compiler had been used, the p4 would have scored 0.144 FLOPs/clock cycle*, which is much better then the g5.

(*which is exactly the same thing as MFLOPs/Mhz. And they wonder why NASA has so much trouble converting things...)

Re:fortran compiler

[Commutative Monoid]

It seems quite possible that Apple wanted vindication for their benchmark "fiasco" (perhaps too strong a word, so forgive me), and this third-party (NASA) with a previous investment in the G4 (the super computer!) might have produced results they favor. I'm not sure I would say that these are nearly as biased (they didn't go out of the way to make the G5 score better this time), and perhaps it's entirely possible they only have licenses for that particular FORTRAN compiler. I don't know, really. There may be sufficient reason to criticize the methodology of the testing for anyone naive enough to consider this an effective comparison of the P4 to the G5 (neither compiler was necessarily the best for its platform), but I personally don't see sufficient evidence to say that the people responsible acted with any intentional bias, unlike what is routinely seen from Apple marketing campaigns.

I realize that Slashdot is in perpetual advertisement for Apple, but I think the userbase is best served by waiting until large numbers of people can provide benchmarks for comparison. Indulging in replying to the trolling that the editors provide is just a waste of energy at this point.